Integrating Nonlinear Independent Component Analysis and Neural Network in Stock Price Prediction

In this research, a time series prediction model by integrating nonlinear independent component analysis (NLICA) and neural network is proposed for stock price. NLICA is a novel feature extraction technique to find independent sources given only observed data that are mixtures of the unknown sources, without prior knowledge of the mixing mechanisms. It assumes that the observed mixtures are the nonlinear combination of latent source signals. The proposed method first use NLICA as preprocessing to transform the input space composed of original time series data into the feature space consisting of independent components (ICs) representing underlying information/features of the original data. Then, the ICs are served as the input variables of the backpropagation neural network (BPN) to build prediction model. Experimental results on Nikkei 225 closing cash index show that the proposed model outperforms the integrated linear ICA and BPN model and the single BPN model without ICA preprocessing.